WARNING: neo4j-import is deprecated
and support
for it will be removed
in a future
version of Neo4j; please use neo4j-admin import instead.
Neo4j
version: 3.2.
1
Importing the contents of these files into
njaq:
Nodes:
/home/dawn/csv/perosnInfo.csv
Relationships:
/home/dawn/csv/know.csv
Available
resources:
Total machine
memory: 3.84 GB
Free machine
memory: 1.61 GB
Max heap memory :
875.00 MB
Processors: 4
Configured max
memory: 700.35 MB
Nodes, started
2017-
06-08
05:35:30.741+
0000
[>
:18.87 MB
|NODE:152.59 MB----|*PROPERTIES(
3)============
|LABEL SCAN--|v:37.14 MB/s-----------]
20.0M ?
21.8K
Done
in 51s
548ms
Prepare node index, started
2017-
06-08
05:36:22.495+
0000
[*
DETECT:419.62 MB----------------------------------------------------------------------------]
20.0M ?-
6500000
Done
in 9s
126ms
Relationships, started
2017-
06-08
05:36:31.678+
0000
[>
:7|T|*PREPARE(
4)=========================================================
|RE|CALCULATE-
|P|v:]
79.9M ?
10.9K
Done
in 4m
17s
742ms
Relationship --> Relationship
1/
1, started
2017-
06-08
05:40:49.548+
0000
[*>-----------------------------------------------------------------------
|LINK------------|v:]
79.9M ?
405K
Done
in 2m
5s
784ms
RelationshipGroup
1/
1, started
2017-
06-08
05:42:55.404+
0000
[*>
:??----------------------------------------------------------------------------------------]
0 ?
0
Done
in 11ms
Node --> Relationship, started
2017-
06-08
05:42:55.439+
0000
[>
:13|*>-------------------------------------------------|LIN
|v:26.00 MB/s--------------------]19.9M ?2.18M
Done in 11s 833ms
Relationship <-- Relationship 1/1, started 2017-06-08 05:43:07.308+0000
[*>-------------------------------------------------------------------------------|LINK----
|v:]79.9M ? 168K
Done in 11m 29s 787ms
Count groups, started 2017-06-08 05:54:37.570+0000
[*>:??----------------------------------------------------------------------------------------] 0 ? 0
Done in 1ms
Gather, started 2017-06-08 05:54:38.061+0000
[*>:??----------------------------------------------------------------------------------------] 0 ? 0
Done in 4ms
Write, started 2017-06-08 05:54:38.156+0000
[*>:??----------------------------------------------------------------------------------------] 0 ? 0
Done in 15ms
Node --> Group, started 2017-06-08 05:54:38.213+0000
[*>:??----------------------------------------------------------------------------------------] 0 ? 0
Done in
Node counts, started 2017-06-08 05:54:38.264+0000
[*>(4)====================================================================================|COU]
20.0M ?
80.0K
Done
in 1m
26s
338ms
Relationship counts, started
2017-
06-08
05:56:04.
625+
0000
[*>(
4)======================================================
|COUNT----------------------------]80.0M ?1.81M
Done in 2m 47s 277ms
IMPORT DONE in 23m 22s 420ms.
Imported:
20000000 nodes
79994052 relationships
80000000 properties
Peak memory usage: 899.62 MB
Neo4J使用导入方法之后会建立索引，否则基本没有性能，建立索引很快。
图算法支持
ArangoDB图算法支持
- AQL
- 遍历：从指定开始点，通过一定算法、边类型、图类型、深度获取与指定开始点相关连通的点。
- 数据源：图、边集合
- 边方向：出边、入边、全部
- 遍历方式：BFS Or DFS
- 最短路径：两点最短路径，选项基本和上面类型
- Pregel
-
@arangodb/pregel 文件夹下，很多分布式的图算法
- PageRank
- CC 强弱连通算法
- 单源最短路径算法
JS扩展
- 通过JS可以完成对内置算法的扩展，但是自定义方法是单线程JS函数，如果用来做算法，性能堪忧，最佳选择就是选择内置的方法去实现图算法。
- 通过JS可以实现很多算法，但是在ArangoSH下代码单线程运作，虽然arangod的JS是在多线程中运行的，但是arangosh是在单线程中运行的，且JS本身并不擅长处理计算型代码，相比之下通过内置的数据库语言而不是这种内置语言与JS混杂方式的代码会快很多；比如Neo4J，OrientDB的查询语言。
Neo4J
对于普通的遍历最短路径算法支持和ArangoDB一样都支持，但Neo4J的图的遍历深度的阈值设置比较难，且深度超过6算法会效率比较低。
相比之下，ArangoDB的算法参数设置全部依赖于Key-Value实现，算法在编码层次灵活性很高。
对于PageRank，CC等算法的实现，Neo4J提供两种方式：
- 编写Jar包，GitHub上有个未被官方承认的Jar包
- Cypher直接实现
OrientDB
同上，图算法也支持Jar包导入。
内置图算法
最短路径
Cypher:
match (p1:person{no:‘%s‘}),(p2:person{no:‘%s‘}) match p=shortestPath((p1)-[*..3]->(p2)) return p
OrientDB SQL:
select dijkstra((select @RID from persons where no=‘%s‘),(select @RID from persons where no=‘%s‘),‘E‘)
AQL:
for v,e in outbound shortest_path ‘%s‘ to ‘%s‘ graph ‘graphPersons‘ return [v._key,e._key]
邻接点
Cypher:
MATCH (js:person)-[:know]-(surfer) WHERE js.no = ‘%s‘ return surfer
OrientDB SQL:
select from E where out = (select @RID from persons where no=‘%s
AQL:
traversal_results = graphPersons.traverse(
start_vertex=‘persons/‘+getSingleInfo(id).no,
strategy=‘bfs‘,
direction=‘outbound‘,
edge_uniqueness=‘global‘,
vertex_uniqueness=‘global‘,
max_depth=1
)
参考资料
- 附录 B. 常见问题_Neo4J最新版本
- ArangoDB 3.0 – A Solid Ground to Scale
- OrientDB calculate PageRank
